Re: [bitbake-devel] [PATCH] bitbake-setup: pass install-buildtools options from config and CLI

Alexander Kanavin <[email protected]> Mon, 15 Jun 2026 17:38:08 +0200
Newsgroups org.openembedded.lists.bitbake-devel
Message-ID <CANNYZj8FzL095jg5nhOqP32hhYj+_6__rS6DGazqtpHmt0yorQ@mail.gmail.com>
On Mon, 15 Jun 2026 at 16:59, Jaipaul Cheernam
<[email protected]> wrote:
> The archive itself contains the same tools as upstream extended buildtools — it can't be provided via recipes because buildtools exist specifically to
> bootstrap the host environment before BitBake can run.
>
> The --url and --filename options are needed for projects that host their buildtools on an internal server rather than yoctoproject.org, and want to pin a
> specific version across the team without requiring manual arguments on every invocation.
>
> The workaround today is to pass arguments manually or modify wrapper scripts, which is probably why this hasn't surfaced as a bug report before.
>
> The --sdk-name option itself is introduced in a companion patch to oe-scripts/install-buildtools in openembedded-core:
> https://lists.openembedded.org/g/openembedded-core/topic/patch_install_buildtools/119814299

Thanks. I'm broadly in favor of this, but I don't know what the
bitbake patch reviewers will say, and especially RP :) Buildtools can
be used to do isolated, reproducible 'container builds' without actual
containers, and I like that.

I wonder if the archive should be downloaded by bitbake-setup itself
using bitbake fetchers (similar to how layers are obtained), and then
given to install-buildtools as a local file, or even executed
directly. There are benefits to this: it will be preserved in the
download cache together with the layers and upstream components (which
helps if upstream server isn't available or altogether disappears),
and checksummed against sha256sum provided in the config (which adds a
reproducibility check, and protection against supply chain attacks,
similar to tarball recipes). What you think?

Alex